Sign in

bearseascape.bsky.social

@bearseascape.bsky.social
6 followers 5 following 17 posts
PostsRepliesMedia
bearseascape.bsky.social @bearseascape.bsky.social · 22/10/2025
See the paper and code here! Paper: arxiv.org/abs/2506.02132 Code: github.com/ml5885/model...
010
bearseascape.bsky.social @bearseascape.bsky.social · 22/10/2025
Takeaways: Despite massive advances in scale, architecture, & training - from BERT to Llama-3.1 - transformer models organize linguistic information the same. The classical NLP pipeline isn't an artifact of early models; it persists in modern ones & seems fundamental to NTP.
110
bearseascape.bsky.social @bearseascape.bsky.social · 22/10/2025
We also looked at attention head outputs vs the residual stream, and find that linguistic information lives primarily in the residual stream. Read the paper for our full analysis including steering vectors, intrinsic dimensionality, and training dynamics (in OLMo2 and Pythia)!
110
bearseascape.bsky.social @bearseascape.bsky.social · 22/10/2025
The probes show that lexical info concentrates in early layers and becomes increasingly nonlinear deeper in the network. Inflectional info remains linearly accessible throughout all layers.
110
bearseascape.bsky.social @bearseascape.bsky.social · 22/10/2025
After confirming that modern LMs rediscover the classical NLP pipeline, we investigated how they encode lemma (word meaning) vs inflectional morphology (word form). We looked at 6 typologically diverse languages: English, Chinese, German, French, Russian and Turkish.
110
bearseascape.bsky.social @bearseascape.bsky.social · 22/10/2025
But there's an important difference: 📊 Large modern models compress the entire pipeline into fewer layers. Information emerges earlier and consolidates faster. Same hierarchical organization, but more efficient. Models seem to have gotten better at building useful representations quickly.
110
bearseascape.bsky.social @bearseascape.bsky.social · 22/10/2025
The hierarchy is remarkably stable across all models: Early layers capture syntax (POS, dependencies) -> Middle layers handle semantics & entities (NER, SRL) -> Later layers encode discourse (coreference, relations) This holds whether you're looking at BERT, Qwen2.5 or OLMo 2.
110
bearseascape.bsky.social @bearseascape.bsky.social · 22/10/2025
🔬We probed 25 models, across 8 linguistic tasks spanning syntax, semantics, and discourse. 🔍The answer: YES. Modern LMs consistently rediscover the classical NLP pipeline.
110
bearseascape.bsky.social @bearseascape.bsky.social · 22/10/2025
🚨 New paper w/ @nsubramani23.bsky.social ! In 2019, Tenney et al. found that BERT rediscovers the classical NLP pipeline, offering linguistic insights into early LMs. But models today are >>100x larger, trained on vastly more data… Do they still rediscover the classical NLP pipeline? 🧵
120
bearseascape.bsky.social @bearseascape.bsky.social · 04/06/2025
🎯 These patterns hold across all 16 models - despite huge differences (encoder/decoder, 100M->8B params, instruction-tuning) Despite rapid advances since BERT, certain aspects of how LMs process language remain remarkably consistent💡 Paper: arxiv.org/abs/2506.02132 Code: github.com/ml5885/model...
arxiv.org
Model Internal Sleuthing: Finding Lexical Identity and Inflectional Morphology in Modern Language Models
Large transformer-based language models dominate modern NLP, yet our understanding of how they encode linguistic information is rooted in studies of early models like BERT and GPT-2. To better underst...
010
bearseascape.bsky.social @bearseascape.bsky.social · 04/06/2025
We also tested how tokenization affects linguistic representations using analogy tasks (king - man + woman = ?) 👑 Whole-word embeddings consistently outperform averaged subtoken representations - linguistic regularities are stored at the word level, not compositionally!
110
bearseascape.bsky.social @bearseascape.bsky.social · 04/06/2025
🔬 We also measured intrinsic dimensionality across layers using PCA. 🎢 Some models (GPT-2, OLMo-2) compress their middle layers to just 1-2 dimensions capturing 50-99% of variance, then expand again! This bottleneck aligns with where grammar is most accessible & lexical info is most nonlinear.
110
bearseascape.bsky.social @bearseascape.bsky.social · 04/06/2025
To understand when these patterns emerge, we analyze OLMo-2 & Pythia checkpoints throughout pre-training. 👶👦👨👨‍🦳 We find that models learn this linguistic organization in the first few thousand steps! But this encoding slowly degrades as training progresses. 📉
110
bearseascape.bsky.social @bearseascape.bsky.social · 04/06/2025
🤔 But are classifiers actually learning linguistic patterns or just memorizing? 📈 We ran control tasks with random labels - inflection classifiers show high selectivity (real learning!) while lemma classifiers don't (memorization).
110
bearseascape.bsky.social @bearseascape.bsky.social · 04/06/2025
Key findings 📊: - 📉 Lexical info concentrates in early layers & becomes increasingly nonlinear in deeper layers - ✨ Inflection (grammar) stays linearly accessible throughout ALL layers - Models memorize word identity but learn generalizable patterns for inflections!
110
bearseascape.bsky.social @bearseascape.bsky.social · 04/06/2025
🧐 How do modern LMs encode linguistic information? Do they represent words grouped by meaning (walk/walked) or grammar (walked/jumped)? We trained classifiers on hidden activations from 16 models (BERT -> Llama 3.1) to find out how they store word identity (lexemes) vs. grammar (inflections).
110
bearseascape.bsky.social @bearseascape.bsky.social · 04/06/2025
🚨New #interpretability paper with @nsubramani23.bsky.social: 🕵️ Model Internal Sleuthing: Finding Lexical Identity and Inflectional Morphology in Modern Language Models
111